AI Agent with Google Gemini

A Python-based AI agent that interacts with files in a sandboxed environment using Google's Gemini API. This project is based on a Boot.dev guided project, enhanced with proper type definitions from the genai documentation for more consistent and reliable results.

Features

• File system operations (read, write, list files)
• Python script execution with arguments
• Sandboxed environment for security
• Iterative processing with function calling
• Mathematical expression evaluation via calculator module
• Verbose mode for detailed operation tracking

Prerequisites

• Python 3.13+
• Google Gemini API key
• uv package manager

Installation

1. Clone the repository:

git clone https://github.com/danlabrador/bootdev-py-ai_agent.git
cd bootdev-py-ai_agent

2. Install dependencies:

uv sync

3. Set up environment variables:

cp .env.example .env
# Add your GEMINI_API_KEY to the .env file

Usage

Running the AI Agent

# Basic usage
uv run main.py "your prompt here"

# Verbose mode for detailed output
uv run main.py "your prompt here" --verbose

Examples

# List all files in the calculator directory
uv run main.py "list all files in the calculator directory"

# Run calculator tests
uv run main.py "run the calculator tests"

# Evaluate a mathematical expression
uv run main.py "use the calculator module we have to calculate 3 * 4 + 5

Project Structure

.
|-- main.py                 # Main agent loop implementation
|-- calculator/             # Example calculator application
|   |-- main.py
|   |-- tests.py
|   |-- pkg/
|       |-- calculator.py  # Calculator class with expression evaluation
|       |-- render.py
|-- functions/              # Agent function implementations
    |-- call_function.py   # Function dispatcher
    |-- get_file_content.py
    |-- get_files_info.py
    |-- run_python_file.py
    |-- write_file.py

Security

The agent implements several security measures:

• All file operations are restricted to relative paths within the working directory
• Path traversal attempts are blocked
• Maximum of 20 iterations per agent session to prevent infinite loops
• Sandboxed execution environment

Configuration

The agent uses the following configuration:

• Model: gemini-2.0-flash-001
• Maximum iterations: 20
• Available functions: file operations (read, write, list) and Python execution

Author

Dan Labrador Email: github.community@danlabrador.com GitHub: @danlabrador

Acknowledgments

• This project is based on a guided project from Boot.dev
• Enhanced with proper type definitions from the official Google genai documentation

License

This project is licensed under the MIT License - see the LICENSE file for details.